Mechanisms and Counterfactuals: a Different Glimpse of the (Secret?) Connexion

Ever since Wesley Salmon’s theory, the mechanical approach to causality has found an increasing number of supporters who have developed it in different directions. Mechanical views such as those advanced by Stuart Glennan, Jim Bogen and Peter Machamer, Lindley Darden and Carl Craver have met with broad consensus in recent years. This paper analyses the main features of these mechanical positions and some of the major problems they still face, referring to the latest debate on mechanisms, causal explanation and the relationship between mechanisms and counterfactuals. I shall claim that the mechanical approach can be recognised as having a fundamental explanatory power, whereas the counterfactual approach, recently developed mainly by Jim Woodw ard and essentially linked to the notion of intervention, has an important heuristic role. Claiming that mechanisms are by no means to be seen as parasitic on counterfactuals or less fundamental than them – as it has been recently suggested –, and that yet counterfactuals can play a part in a conceptual analysis of causation, I shall look for hin ts in support of the peaceful coexistence of the two. 1. The causal structure of the world: processes and interactions, entities and activities In the last couple of decades philosophy of science has seen the elaboration of several mechanical accounts. While terminology and emphasis differ, all the theories overlap in believing that mechanisms are complex systems present in nature. Wesley Salmon’s philosophical work is unanimously regarded as the compulsory locus for anyone interested in the notion of mechanism since the eighties. As is well-known, Salmon has developed a “process theory” of causation, centred on the notions of causal process, causal RAFAELLA CAMPAN ER 16 1 Apart from Craver (2001:69-70). production and causal propagation. In short, causal processes are defined as spatio-temporally continuous processes which exhibit consistency of structure over time, and are capable of transmitting a modification of their structure from the point at which it is performed onwards, without additional interventions. The production of causal influence is accounted for by appealing to causal forks, characterised in statistical terms. Once produced, causal influence is propagated continuously through processes. Interacting processes constitute a mechanical, objective and probabilistic network, underlying phenomena and responsible for their occurrence. Salmon’s conception of causality goes hand in hand with his theory of causal explanation, which comprises two levels: we first need to identify the properties which are statistically relevant with respect to the occurrence of the event to be explained; we then account for them in terms of the net of causal processes underlying the event. A further distinction Salmon makes, not recalled by later authors, is that between etiological and constitutive causation. When we aim at explaining a given event E, we may look at E as occupying a finite volume of four-dimensional space-time. If we want to show why E occurred, we fill in the causally relevant processes and interactions that occupy the past light cone of E. This is the etiological aspect of our explanation; it exhibits E as embedded in its causal nexus. If we want to show why E manifests certain characteristics, we place inside the volume occupied by E the internal causal mechanisms that account for E’s nature. This is the constitutive aspect of our explanation; it lays bare the causal structure of E. (Salmon, 1984:275) (emphasis added) In sum, causal-mechanical explanations of the etiological sort illustrate the causal story leading up to the occurrence of the explanandum, whereas constitutive explanations provide a causal analysis by showing the underlying causal mechanisms that constitute the phenomenon itself. On the whole, Salmon’s conception aims at conjugating mechanicism and probability, both indispensable for an adequate picture of the causal structure of the world. Neither statistical relevance relations nor MECHANISMS AND COUNTERFACTU ALS 17 connecting causal processes have explanatory import if taken on their own; they do so only together. In the early nineties, Phil Dowe, among others, has largely criticised Salmon’s view, raising objections against its being circular, using vague terms (such as “structure”), characterizing causal production and causal interactions in terms of statistical relations and, last but not least, using counterfactuals in the formulation of the criteria defining causal processes and causal interaction. According to Salmon, a causal process is such that, had a modification of its structure been performed, it would have transmitted it from that point onwards; a causal interaction is an intersection between two causal processes such that, had they intersected, both their structures would have been modified from that point onwards. Dowe has thus advanced a new process theory, called “the conserved-quantity theory”, which aims at preserving Salmon’s objective and physical idea of causation, while getting rid of counterfactuals. In short, the conserved-quantity theory, embraced with minor modifications by Salmon himself, holds that a causal process is the world line of an object exhibiting a conserved quantity, and a causal interaction is an intersection of processes in which an exchange of a conserved quantity occurs. The next attempt to elaborate a mechanical position, not intended as a direct further development of Salmon-Dowe’s view, has been made by Stuart Glennan roughly in the same years. Glennan wants to substitute Salmon’s and Dowe’s process causal-mechanical theories with what he calls a “complex-systems account”. Its core is the following definition: A mechanism for a behaviour is a complex system that produced that behaviour by the interaction of a number of parts, where the interactions between parts can be characterised by direct, invariant, change-relating generalizations. (Glennan, 2002:S344) The notion of mechanism is here strongly linked to that of behaviour: Glennan believes that a mechanism cannot even be identified without mentioning what it does. As in Salmon’s view, a central role is played by the notion of interaction, though this is not as precisely defined. According to Glennan, no a priori restrictions are to be put on the sorts of allowable interactions that may take place between the parts of a mechanism. Whereas “interaction” means something very specific and RAFAELLA CAMPAN ER 18 2 Glennan appeals to such a no tion in his (2002). In the definition given in his (1996) mechanisms are claimed to be working “according to direct causal laws” (pp. 50). circumscribed for Salmon and Dowe, Glennan’s account takes the relevant modes of interaction between the component parts of mechanisms to always depend upon the behaviour we are interested in explaining. Mechanisms must simply be such that their “‘internal’ parts interact to produce a system’s ‘external’ behaviour” (Glennan, 1996:49), but it is far from clear how we shall make sense of “internal” as opposed to “external”, and what can properly count as “parts” of a mechanism. Glennan’s view of mechanical causation is meant to be a theory of causal explanation too. Mechanisms are made up of parts, and events are claimed to be causally related when there is a mechanism that connects them; a good description of a mechanism is believed to provide an adequate causal explanation. As emerges from the key-definition quoted above, the interactions between parts of the mechanism which give rise to its behaviour are characterised by invariant generalizations. Glennan admittedly borrows this notion from Jim Woodward, and takes it to indicate a generalization that would hold were a range of possible interventions to be performed. According to Glennan, a two-way relationship holds between invariant generalizations and mechanisms: First, reliable behaviour of mechanisms depends upon the existence of invariant relations between their parts, and changerelating generalizations characterise these relations. Second, many such generalizations are mechanically explicable, in the sense that they are just generalizations about the behaviour of mechanisms. A single generalization can both be explained by a mechanism and characterise the interaction between parts of a larger mechanism. (Glennan, 2005:445-446). The link between the notions of mechanism and of invariant generalization turns out to be a very strict one. Being complex, or often very complex, systems, mechanisms can be decomposed into subsystems. Decomposition depends on what is being explained, but, Glennan warns, its context-dependence must not be confused with antirealism or relativism: descriptions of mechanisms are MECHANISMS AND COUNTERFACTU ALS 19 adequate descriptions only insofar as they show what is really there. Whereas Salmon recognised a specific part of physics, that of “spooky” actions at-a-distance studied in quantum mechanics, as the only deeply problematic field for his mechanical theory of causation, for Glennan all laws but the fundamental laws of physics can be explained in mechanical terms. Finally, let me just recall that Glennan’s most recent work (2005) focuses on the nature and testing of mechanical models, where the latter are claimed to consist of both the description of the mechanism’s behaviour and the description of the mechanism accounting for that behaviour. A distinction is thus drawn between the mechanism as such, and our conceptual model aimed at representing it. That what mechanisms do is closely linked to how they do it, namely to how they are organised, is very strongly highlighted by Peter Machamer, Lindley Darden and Carl Craver, whose theory is probably nowadays the most famous and debated mechanical account. The articles that, part jointly and part separately, Machamer, Darden and Craver have written in the last six years present mechanisms as entities a